Towards three-dimensional discrete fracture network modeling using integrated multidimensional outcrop data.

Area of Science:

• Geosciences
• Artificial Intelligence
• Computational Geology

Background:

• Analogue outcrops offer discrete fracture network (DFN) data beyond seismic and well log resolution.
• Fracture characterization requires specialized interpretation of geometric, topological, and geomechanical attributes.
• Outcrop data can be limited, potentially impacting the accuracy of DFN models.

Purpose of the Study:

• To develop an AI-driven methodology for fracture characterization and DFN modeling using digital outcrop data.
• To mitigate bias in DFN analysis and parameter estimation through artificial intelligence.
• To create deterministic and stochastic volumetric models incorporating fracture persistence.

Main Methods:

• Spherical clustering with a modified K-Means algorithm for 3D fracture data (strike/dip).
• Stochastic Gradient Descent for power-law estimation, considering fractal fracture propagation.
• Application to Jandaíra formation carbonate rocks in the Potiguar basin for case study.

Main Results:

• AI integration successfully mitigated bias in DFN analysis and parameter estimation.
• Deterministic and stochastic volumetric models with fracture persistence were generated.
• Stronger correlation between 2D and 3D data was observed for connectivity, fracture intensity, and length distribution.

Conclusions:

• The proposed AI methodology enhances the accuracy of DFN modeling from digital outcrop data.
• The approach effectively integrates diverse fracture attributes for robust model generation.
• AI-driven fracture characterization provides a more reliable basis for subsurface geological modeling.